Don't hold your breath, Tech Insiders. Apple's working to rein in overconfident AI, Microsoft's custom chips are running late, and Tesla's robotaxis can't seem to drive themselves. Automation may be the future, but this week, it's having a very human moment. Let's take a look at what's still buffering. |
|
|
Here's what you need to know today: |
|
|
Apple Teaches Siri When to Hit Pause |
Siri's getting smarter… but she still might press the wrong button. Researchers at Apple and the University of Washington have built a "risk taxonomy" for mobile actions, ranking everything from harmless refreshes to irreversible account nukes. They collected real smartphone interactions, labeling each tap according to its cost, reversibility, privacy impact, and potential impact on others. They then fine-tuned large language models, including GPT-4, to predict when an action requires user confirmation. |
Early wins: The taxonomy boosted models' accuracy, but even GPT-4 topped out around 58% correct. This means that bots still confuse "clear history" with "clear savings." The models also over-flag low-stakes taps, which could turn future Siri into a relentless hall monitor unless Apple dials in better context. The takeaway: Mobile AI isn't just about spotting buttons; it's about grasping consequences. Apple's next-gen Siri (rumored for 2026) will need a healthy dose of caution—and maybe a second opinion—before hitting "Delete Account." Why it matters: Your phone holds everything from gym selfies to mortgage apps. An overeager assistant could wreak havoc or treat every swipe as if it were defusing a bomb. The study shows that today's models still over-flag harmless taps, so Apple has to teach future Siri the difference between "refresh feed" and "launch nukes"—maybe that's why the big Siri upgrade got punted to 2026. |
|
|
| Would you trust Siri to make purchases for you? |
|
|
Results from Yesterday's Pulse Check |
Which CC Signal would you turn on first? |
|
|
Microsoft's In-House AI Chip Hits Big Delay |
Chip happens—even in Redmond. Microsoft's next-gen Maia processor (code-named Braga) has been delayed at least six months, pushing mass production to 2026. Reports blame surprise design changes, staffing churn, and OpenAI-driven feature tweaks that destabilized early simulations. Worse, insiders say performance will "fall well short" of Nvidia's Blackwell GPUs, the current gold standard Microsoft hoped to dethrone. |
Image Source: ChatGPT (DALL·E) |
The delay hobbles Microsoft's plan to slash its Nvidia bill and run Azure AI on homegrown silicon. Meanwhile, impatient customers aren't waiting for Braga: OpenAI is already trialing Google's seventh-gen TPUs to cut costs, and Amazon's Trainium3 is set to ship later this year. Microsoft now pins its hopes on later variants—Braga-R in 2027 and Clea after that—to close the gap. |
|
|
Kill 2G Now to Block Fake-Cell-Tower Text Attacks |
Police busted crooks using "SMS blasters" that mimic cell towers and beam scam texts straight to phones—sidestepping carrier spam filters. Security experts warn users to kill 2G connectivity, the weak link that powers this exploit. |
We've heard it before: Ditch 2G if you value your data. But with SMS blasters on the rise, it's more urgent than ever. Android 16's Advanced Protection disables it, future Pixel phones may warn you of insecure networks without it, and iPhone owners would need Lockdown Mode. Most US carriers have already shuttered 2G (AT&T and Verizon did so years ago), and T-Mobile has committed to its 2G phase-out in 2025. While 3G networks have also largely been retired, it's best to stick to 4G/5G on your phone. |
Grocery Giant Breach Exposes 2.2 Million Records |
Ahold Delhaize confirmed that November's ransomware attack siphoned data on 2.2 million people—potentially including SSNs, bank account numbers, employment, and medical details—across US grocery chains such as Food Lion and Stop & Shop. The ransomware gang INC Ransom published samples of the stolen files on its dark web leak site. However, the retailer stated that the attackers did not compromise customer-facing payment or pharmacy systems, and no credit card numbers were among the stolen files. Individuals affected should monitor their statements, consider placing credit freezes or fraud alerts, and enroll in any offered identity theft protection. |
Tesla Robotaxi Already Needs a Human Hand |
The fleet counts only a dozen cars within a tiny geofence, yet incidents keep stacking up. Elon Musk still promises 1,000 Austin robotaxis in months and a million nationwide by 2026. Reality, complete with flashing hazard lights, suggests a longer road. When the robot drives like it's taken six espresso shots, be glad a human still has the wheel. |
|
|
| Writer at TechnologyAdvice |
Justin Meyers is an investigative writer and editor who draws on over a decade of meticulous hands-on research to deliver the full, trustworthy story behind consumer and enterprise tech, including cybersecurity. |
|
|
| Writer at TechnologyAdvice |
Justin Meyers is an investigative writer and editor who draws on over a decade of meticulous hands-on research to deliver the full, trustworthy story behind consumer and enterprise tech, including cybersecurity. |
|
|
Curious about where AI is really headed? |
The Neuron cuts through the noise to bring you smart, hype-free takes on the latest AI trends, tools, and breakthroughs. Join 500,000+ professionals from top companies like Microsoft, Apple, Salesforce and more. |
|
|
Advertise in Daily Tech Insider! Daily Tech Insider is a TechnologyAdvice business © 2025 TechnologyAdvice, LLC. All rights reserved. TechnologyAdvice, 3343 Perimeter Hill Dr., Suite 100, Nashville, TN 37211, USA. |
|
|
|
